AGENTS VIC Meeting

Friday, 25 March 2011 @ 9:30AM.
Coffee, tea and refreshments from 9:15.

University of Melbourne, Rm 5.08, ICT Building (Dept of CSSE), 111 Barry St.

Wojciech Lorkiewicz, Swinburne University of Technology

Aligning the meaning of language symbols in a multiagent system

Language is an everyday tool that facilitates interaction and communication between agents and is extensively used to gain, share and utilise information in a social setting. However, to successfully communicate with each other, the agents must align their private and autonomous representations and form a system of shared conventions. As such, the superior goal is to study different mechanisms of alignment and define the necessary settings for resultant formulation of common substance of language. In this research, word learning is perceived as a task of mapping labels onto a pre-established set of concept, and the aim is to investigate current models, and design new ones, assuring consistency in the identified basic scenarios of language alignment task, each representing a possible case of language utilisation in a multi-agent system. In this talk we follow the Language Game Model (an intuitive scheme of interaction proposed by Steels) and investigate how a population of communicating agents can emerge a shared lexicon and reach lexical conventions.

Wilson Wong, RMIT University

The Web as a Source of Knowledge for Open Domain Conversational Agents

Existing conversational agents (CAs) for general chats as well as goal-oriented interactions do not have the capacity for long-term relationship building. Existing CAs are unable to perform all of the functions crucial for realistic and engaging conversations such as predicting interest and intention, ensuring coherence and engagement, and automatically updating knowledge. This talk focuses on the potential use of emergent semantics on the Web for deriving and updating knowledge for conversational agents.